作者: Maria De Marsico , Luca Moschella , Andrea Sterbini , Marco Temperini
DOI: 10.1007/978-3-319-66610-5_31
关键词: Topology (electrical circuits) 、 Machine learning 、 Global network 、 Peer assessment 、 Artificial intelligence 、 Variable-order Bayesian network 、 Bayesian network 、 Theoretical computer science 、 Selection (relational algebra) 、 Network topology 、 Computer science 、 Bayesian inference
摘要: The paper investigates if and how the topology of peer-assessment network can affect performance Bayesian model adopted in OpenAnswer. Performance is evaluated terms comparison predicted grades with actual teacher’s grades. global built by interconnecting smaller subnetworks, one for each student, where intra-subnetwork nodes represent student’s characteristics, peer assessment assignments make up inter-subnetwork connections determine evidence propagation. A possible subset teacher graded answers dynamically determined suitable selection stop rules. research questions addressed are: (RQ1) “does (diameter) negatively influence precision grades?”; affirmative case, (RQ2) “are we able to reduce negative effects high-diameter networks through an appropriate choice students be corrected teacher?” We show that OpenAnswer less effective on higher diameter topologies, this avoided chosen considering topology.